7 Comments

Building on your general argument. This is a useful post from a16z https://a16z.com/the-empty-promise-of-data-moats/

Expand full comment

Also, purpose-built UIs can be a moat.

Expand full comment

There's a huge gap in what is official and what happens in reality. I'm in the legal space and it's clear that there's no massive documentation repository or data set containing the advice lent to clients because that advice does NOT always match a logical deduction based on the rules. Some professions have an enormous store of personal experience and poorly documented data that plays a key role in their success.

Expand full comment

There is a lot of ai training data from online content, journals, articles, enterprises and such but do you feel like there is a gap in ai training data somewhere?

Expand full comment

Completely agree... Unfortunately most of the organaisations running behind THE BEST models instead of managing better data quality...

Expand full comment

Have a look at Cynthia Systems and one of our recent blog posts:

Paving the Way for Bias-Free Artificial Intelligence

https://cynthiasystems.com/blog/paving-the-way.html

Cynthia has a powerful algorithmic moat, and it is highly unlikely that competitors such as Algolia and Bloomreach will reinvent it anytime soon, as their value setting is at odds with the discovery.

Additonally, no LLM such as GPT-4 can achieve this use case as an entire product catalog, or an entire music streaming app and associated playlists, can be loaded into a context window, let alone queried in real time with guaranteed correctness.

Expand full comment

We don’t have moats, but we may have a garbage collector of all the stupid things people want our llm to do?

Expand full comment